Optimal portfolio in partially observed stochastic volatility models. (Q1872462)

The authors study an incomplete financial model with one bond and \(n\) risky assets whose price process \((S_t)\) has a dynamics governed by \[ dS_t= \mu_t dt+ \sigma(t, S_t, Y_t)\,dW_t,\quad 0\leq t\leq T. \] The stochastic volatility \(\sigma(\cdot,\cdot,\cdot)\) is influenced by some latent \(\mathbb{R}^d\)-valued process \((Y_t)\) satisfying \[ dY_t= \eta_t dt+ \rho(t, S_t, Y_t)\,dW_t+ \gamma(t, S_t, Y_t)\,dB_t,\quad 0\leq t\leq T. \] Here, \((W_t)\) and \((B_t)\) are independent Brownian motions, and the drifts \((\mu_t)\), \((\eta_t)\) are adapted processes. Suppose that an investor can only observe the stock prices. Using stochastic filtering techniques and adapting martingale duality methods, the authors solve the portfolio optimization problem when the investor wants to maximize his expected utility from terminal wealth. Furthermore, the Bayesian case is studied when the risk premia of the stochastic volatility model are unobservable random variables with known prior distribution.